Sonographic diagnosis of thyroid cancer with support of AI

Machine learning for risk stratification of thyroid cancer patients: a 15-year cohort study

PURPOSE

The objective of this study was to train machine learning models for predicting the likelihood of recurrence in patients diagnosed with well-differentiated thyroid cancer. While thyroid cancer mortality remains low, the risk of recurrence is a significant concern. Identifying individual patient recurrence risk is crucial for guiding subsequent management and follow-ups.

METHODS

In this prospective study, a cohort of 383 patients was observed for a minimum duration of 10 years within a 15-year timeframe. Thirteen clinicopathologic features were assessed to predict recurrence potential. Classic (K-nearest neighbors, support vector machines (SVM), tree-based models) and artificial neural networks (ANN) were trained on three distinct combinations of features: a data set with all features excluding American Thyroid Association (ATA) risk score (12 features), another with ATA risk alone, and a third with all features combined (13 features). 283 patients were allocated for the training process, and 100 patients were reserved for the validation of stage.

RESULTS

The patients' mean age was 40.87 ± 15.13 years, with a majority being female (81%). When using the full data set for training, the models showed the following sensitivity, specificity and AUC, respectively: SVM (99.33%, 97.14%, 99.71), K-nearest neighbors (83%, 97.14%, 98.44), Decision Tree (87%, 100%, 99.35), Random Forest (99.66%, 94.28%, 99.38), ANN (96.6%, 95.71%, 99.64). Eliminating ATA risk data increased models specificity but decreased sensitivity. Conversely, training exclusively on ATA risk data had the opposite effect.

CONCLUSIONS

Machine learning models, including classical and neural networks, efficiently stratify the risk of recurrence in patients with well-differentiated thyroid cancer. This can aid in tailoring treatment intensity and determining appropriate follow-up intervals.

Machine learning-aided metallomic profiling in serum and urine of thyroid cancer patients and its environmental implications

The incidence rate of thyroid cancer has been growing worldwide. Thyroid health is closely related with multiple trace metals, and the nutrients are essential in maintaining thyroid function while the contaminants can disturb thyroid morphology and homeostasis. In this study, we conducted metallomic analysis in thyroid cancer patients (n = 40) and control subjects (n = 40) recruited in Shenzhen, China with a high incidence of thyroid cancer. We found significant alterations in serumal and urinary metallomic profiling (including Cr, Mn, Fe, Co, Ni, Cu, Zn, As, Sr, Cd, I, Ba, Tl, and Pb) and elemental correlative patterns between thyroid cancer patients and controls. Additionally, we also measured the serum Cu isotopic composition and found a multifaceted disturbance in Cu metabolism in thyroid disease patients. Based on the metallome variations, we built and assessed the thyroid cancer-predictive performance of seven machine learning algorithms. Among them, the Random Forest model performed the best with the accuracy of 1.000, 0.858, and 0.813 on the training, 5-fold cross-validation, and test set, respectively. The high performance of machine learning has demonstrated the great promise of metallomic analysis in the identification of thyroid cancer. Then, the Shapley Additive exPlanations approach was used to further interpret the variable contributions of the model and it showed that serum Pb contributed the most in the identification process. To the best of our knowledge, this is the first study that combines machine learning and metallome data for cancer identification, and it supports the indication of environmental heavy metal-related thyroid cancer etiology.

A new xAI framework with feature explainability for tumors decision-making in Ultrasound data: comparing with Grad-CAM

BACKGROUND AND OBJECTIVE

The value of implementing artificial intelligence (AI) on ultrasound screening for thyroid cancer has been acknowledged, with numerous early studies confirming AI might help physicians acquire more accurate diagnoses. However, the black box nature of AI's decision-making process makes it difficult for users to grasp the foundation of AI's predictions. Furthermore, explainability is not only related to AI performance, but also responsibility and risk in medical diagnosis. In this paper, we offer Explainer, an intrinsically explainable framework that can categorize images and create heatmaps highlighting the regions on which its prediction is based.

METHODS

A dataset of 19341 thyroid ultrasound images with pathological results and physician-annotated TI-RADS features is used to train and test the robustness of the proposed framework. Then we conducted a benign-malignant classification study to determine whether physicians perform better with the assistance of an explainer than they do alone or with Gradient-weighted Class Activation Mapping (Grad-CAM).

RESULTS

Reader studies show that the Explainer can achieve a more accurate diagnosis while explaining heatmaps, and that physicians' performances are improved when assisted by the Explainer. Case study results confirm that the Explainer is capable of locating more reasonable and feature-related regions than the Grad-CAM.

CONCLUSIONS

The Explainer offers physicians a tool to understand the basis of AI predictions and evaluate their reliability, which has the potential to unbox the "black box" of medical imaging AI.

Artificial Intelligence in Thyroid Field-A Comprehensive Review

Simple Summary

The incidence of thyroid pathologies has been increasing worldwide. Historically, the detection of thyroid neoplasms relies on medical imaging analysis, depending mainly on the experience of clinicians. The advent of artificial intelligence (AI) techniques led to a remarkable progress in image-recognition tasks. AI represents a powerful tool that may facilitate understanding of thyroid pathologies, but actually, the diagnostic accuracy is uncertain. This article aims to provide an overview of the basic aspects, limitations and open issues of the AI methods applied to thyroid images. Medical experts should be familiar with the workflow of AI techniques in order to avoid misleading outcomes.

Abstract

Artificial intelligence (AI) uses mathematical algorithms to perform tasks that require human cognitive abilities. AI-based methodologies, e.g., machine learning and deep learning, as well as the recently developed research field of radiomics have noticeable potential to transform medical diagnostics. AI-based techniques applied to medical imaging allow to detect biological abnormalities, to diagnostic neoplasms or to predict the response to treatment. Nonetheless, the diagnostic accuracy of these methods is still a matter of debate. In this article, we first illustrate the key concepts and workflow characteristics of machine learning, deep learning and radiomics. We outline considerations regarding data input requirements, differences among these methodologies and their limitations. Subsequently, a concise overview is presented regarding the application of AI methods to the evaluation of thyroid images. We developed a critical discussion concerning limits and open challenges that should be addressed before the translation of AI techniques to the broad clinical use. Clarification of the pitfalls of AI-based techniques results crucial in order to ensure the optimal application for each patient.

Novel MRI-Based CAD System for Early Detection of Thyroid Cancer Using Multi-Input CNN

Early detection of thyroid nodules can greatly contribute to the prediction of cancer burdening and the steering of personalized management. We propose a novel multimodal MRI-based computer-aided diagnosis (CAD) system that differentiates malignant from benign thyroid nodules. The proposed CAD is based on a novel convolutional neural network (CNN)-based texture learning architecture. The main contribution of our system is three-fold. Firstly, our system is the first of its kind to combine T2-weighted MRI and apparent diffusion coefficient (ADC) maps using a CNN to model thyroid cancer. Secondly, it learns independent texture features for each input, giving it more advanced capabilities to simultaneously extract complex texture patterns from both modalities. Finally, the proposed system uses multiple channels for each input to combine multiple scans collected into the deep learning process using different values of the configurable diffusion gradient coefficient. Accordingly, the proposed system would enable the learning of more advanced radiomics with an additional advantage of visualizing the texture patterns after learning. We evaluated the proposed system using data collected from a cohort of 49 patients with pathologically proven thyroid nodules. The accuracy of the proposed system has also been compared against recent CNN models as well as multiple machine learning (ML) frameworks that use hand-crafted features. Our system achieved the highest performance among all compared methods with a diagnostic accuracy of 0.87, specificity of 0.97, and sensitivity of 0.69. The results suggest that texture features extracted using deep learning can contribute to the protocols of cancer diagnosis and treatment and can lead to the advancement of precision medicine.

Artificial intelligence in ultrasound

Ultrasound (US), a flexible green imaging modality, is expanding globally as a first-line imaging technique in various clinical fields following with the continual emergence of advanced ultrasonic technologies and the well-established US-based digital health system. Actually, in US practice, qualified physicians should manually collect and visually evaluate images for the detection, identification and monitoring of diseases. The diagnostic performance is inevitably reduced due to the intrinsic property of high operator-dependence from US. In contrast, artificial intelligence (AI) excels at automatically recognizing complex patterns and providing quantitative assessment for imaging data, showing high potential to assist physicians in acquiring more accurate and reproducible results. In this article, we will provide a general understanding of AI, machine learning (ML) and deep learning (DL) technologies; We then review the rapidly growing applications of AI-especially DL technology in the field of US-based on the following anatomical regions: thyroid, breast, abdomen and pelvis, obstetrics heart and blood vessels, musculoskeletal system and other organs by covering image quality control, anatomy localization, object detection, lesion segmentation, and computer-aided diagnosis and prognosis evaluation; Finally, we offer our perspective on the challenges and opportunities for the clinical practice of biomedical AI systems in US.

Segmentation and classification of thyroid follicular neoplasm using cascaded convolutional neural network

In this paper, we present a segmentation and classification method for thyroid follicular neoplasms based on a combination of the prior-based level set method and deep convolutional neural network. The proposed method aims to discriminate thyroid follicular adenoma (TFA) and follicular thyroid carcinoma (FTC) in ultrasound images. In their appearance, these two kinds of tumours have similar shapes, sizes and contrasts. Therefore, it is difficult for even ultrasound specialists to distinguish them. Because of the complex background in thyroid ultrasound images, before distinguishing TFA and FTC, we need to segment the lesions from the whole image for each patient. The main challenge of segmentation is that the images often have weak edges and heterogeneous regions. The main issue of classification is that the accuracy depends on the features extracted from the segmentation results. To solve these problems, we conduct the two tasks, i.e. segmentation and classification, by a cascaded learning architecture. For segmentation, to obtain more accurate results, we exploit the Res-U-net framework and the prior-based level set method to enhance their respective abilities. Then, the classification network is trained by sharing shallow layers of the segmentation network. Testing the proposed method on real patient data shows that it is able to segment the lesion areas in thyroid ultrasound images with a Dice score of 92.65% and to distinguish TFA and FTC with a classification accuracy of 96.00%.

Artificial Intelligence in Medical Imaging and Its Application in Sonography for the Management of Liver Tumor

Recent advancement in artificial intelligence (AI) facilitate the development of AI-powered medical imaging including ultrasonography (US). However, overlooking or misdiagnosis of malignant lesions may result in serious consequences; the introduction of AI to the imaging modalities may be an ideal solution to prevent human error. For the development of AI for medical imaging, it is necessary to understand the characteristics of modalities on the context of task setting, required data sets, suitable AI algorism, and expected performance with clinical impact. Regarding the AI-aided US diagnosis, several attempts have been made to construct an image database and develop an AI-aided diagnosis system in the field of oncology. Regarding the diagnosis of liver tumors using US images, 4- or 5-class classifications, including the discrimination of hepatocellular carcinoma (HCC), metastatic tumors, hemangiomas, liver cysts, and focal nodular hyperplasia, have been reported using AI. Combination of radiomic approach with AI is also becoming a powerful tool for predicting the outcome in patients with HCC after treatment, indicating the potential of AI for applying personalized medical care. However, US images show high heterogeneity because of differences in conditions during the examination, and a variety of imaging parameters may affect the quality of images; such conditions may hamper the development of US-based AI. In this review, we summarized the development of AI in medical images with challenges to task setting, data curation, and focus on the application of AI for the managements of liver tumor, especially for US diagnosis.

AIBx, Artificial Intelligence Model to Risk Stratify Thyroid Nodules

Background: Current classification systems for thyroid nodules are very subjective. Artificial intelligence (AI) algorithms have been used to decrease subjectivity in medical image interpretation. One out of 2 women over the age of 50 years may have a thyroid nodule and at present the only way to exclude malignancy is through invasive procedures for those that are suspicious on ultrasonography. Hence, there exists a need for noninvasive objective classification of thyroid nodules. Some cancers have benign appearance on ultrasonogram. Hence, we decided to create an image similarity algorithm rather than image classification algorithm. Materials and Methods: Ultrasound images of thyroid nodules from patients who underwent either biopsy or thyroid surgery from February 2012 to February 2017 in our institution were used to create AI models. Nodules were excluded if there was no definitive diagnosis of it being benign or malignant. A total of 482 nodules met the inclusion criteria and all available images from these nodules were used to create the AI models. Later, these AI models were used to test 103 thyroid nodules that underwent biopsy or surgery from March 2017 to July 2018. Results: Negative predictive value (NPV) of the image similarity model was 93.2%. Sensitivity, specificity, positive predictive value (PPV), and accuracy of the model were 87.8%, 78.5%, 65.9%, and 81.5%, respectively. Conclusions: When compared with published results of ultrasound thyroid cancer risk stratification systems, our image similarity model had comparable NPV with better sensitivity, specificity, and PPV. By using image similarity AI models, we can decrease subjectivity and decrease the number of unnecessary biopsies. Using image similarity AI model, we were able to create an explainable AI model that increases physician's confidence in the predictions.